The effects of misclassification in studies of gene-environment interactions.

Differential Misclassification of genotype may cause usual tests producing spurious gene-disease associations or unable to detect important associations. To control for the resulting bias, many model-based approaches have been proposed using validation data. However, the effects of joint misclassification of genotype and environmental exposure in studies of gene-environment interaction are still not clear. In this paper, we focus on quantifying the effects of misclassification in case-control and case-only studies of interaction without specifying any model for misclassification probabilities. By using the derived results, we are able to identify important conditions, under which the presence of misclassification does not introduce bias in case-control or case-only studies. We have conducted a simulation study, using parameter values from real examples, to confirm our theoretical findings. The simulation results show that under the identified conditions, many regular tests for the hypothesis of no interaction maintain correct type I error rates in the presence of differential misclassification. However, their powers may be decreased as misclassification error rates increase. Copyright 2008 S. Karger AG, Basel.